Clinical heterogeneity is the hallmark of chronic lymphocytic leukemia (CLL), with some patients surviving for decades without needing treatment, while others present early progression and a short overall survival, if untreated [1]. Therefore, great efforts have been made in order to best predict CLL evolution at the time of diagnosis, and for this, the most widely used markers are cytogenetic parameters [2]. To this end, Dohner’s prognostic algorithm for clinical aggressiveness was proposed by combining conventional chromosome banding with the four main cytogenetic alterations detected by fluorescence in situ hybridization (FISH) since these abnormalities are present in more than 80% of CLL cases [3]. Indeed, patients with deletion (del)17p, del(11q), and a complex karyotype (≥ 3 chromosomal abnormalities) most frequently have an unfavorable outcome (high risk) followed by patients with trisomy 12 or a normal karyotype (intermediate risk). In contrast, over 30% of the cases are represented by patients with an isolated del(13q) and exhibit a favorable evolution (low risk). The main limitation of this cytogenetic approach is related to the fact that cytogenetic abnormalities can evolve during the course of the disease with new subclones which can substitute for those previously established, by presenting either new deletions (e.g., del(17p) or del(11q)) or driver mutations (e.g., TP53, SF3B1), and such an effect can be enhanced by chemotherapy [4].

In parallel, studies have investigated the outcome of CLL patients depending on epigenetic modifications as determined by epigenome-wide association studies (EWAS), genome-wide chromatin accessibility maps, and microRNA analysis showing that loss of epigenetic stability is associated with disease progression [5, 6]. The link between DNA methylation loss in promoters and a higher probability of harboring a subclonal driver mutation was highlighted, the consequence being an adverse clinical outcome [5]. The list of tumor regulator genes and markers with high prognostic significance, like TCL1, CD20, and LPL, which display altered DNA methylation regulation, is continuously growing [7–9]. DNA demethylation also occurs at repetitive elements, suggesting that global DNA demethylation reported in CLL largely accounts for repetitive elements that represent half of the chromatin [10, 11]. The mechanisms underlying aberrant DNA methylation in CLL are complex and do not rely only on the control of DNA methylation by DNA methyl transferases (DNMT). DNA demethylation, catalyzed by the ten-eleven translocation (TET) enzymes, has also been shown to be altered in CLL [12, 13].

The goal of the present work was to ascertain the associations between the main cytogenetic risk factors of the disease and the two cytosine derivatives: 5-methylcytosine (5-mCyt) and 5-hydroxymethylcytosine (5-hmCyt). DNMT and TET transcript levels were also studied by real-time quantitative (RTq)-PCR. The main observation highlighted by this study is that combining a conventional cytogenetic approach with global 5-mCyt assessment drastically improves CLL outcome prediction in patients with del(13q).

Assessment of the cytogenetic status in CLL at diagnosis is routinely performed by using karyotype and FISH analysis for prognostic purposes. However, such an approach poorly integrates clonal evolution, thus indicating that the well-established Dohner’s classification can be improved. Accordingly, we have hypothesized that coupling cytogenetic and epigenetic factors would help predict disease outcome, and in addition, we have confirmed the importance of epigenetic modifications in CLL outcome.

Interplay exists between genetic and epigenetic mechanisms during CLL evolution, and two models were proposed for this co-evolution, one based on a stepwise acquisition model from early clonal mutations/cytogenetic abnormalities (e.g., del(13q)), while in the second model, a simultaneous acquisition was proposed (e.g., del(17p)) [18, 19]. The simultaneous acquisition model is supported by the observation that del(17p) involving the tumor suppressor TP53 is associated with global DNA demethylation in CLL B cells [10, 20] and by our observation that CLL patients from the cytogenetic intermediate- and/or high-risk groups present a reduction in 5-mCyt and 5-hmCyt levels and in DNMT3A and TET1/2 transcripts. In the second model, as observed in CLL patients with low-risk cytogenetic abnormalities such as del(13q), the global DNA methylation pattern is suspected to be relatively stable over time, and when disease progression occurs, selection of novel global DNA methylation patterns evolves independently of the genetic heterogeneity [18, 21]. Indeed, del(13q) is suspected to be an early event as confirmed by the detection of this cytogenetic abnormality in the CD34⁺ cell population in CLL patients [22]. Our study further supports the stepwise dynamic model of clonal populations in CLL with the del(13q) as an early event which can be associated with epigenetic dysregulations as this clonal population expands. This is highlighted by the observation that CLL patients with a higher percentage (above 80%) of 13q deleted nuclei also have a defective DNA methylation process and a shorter PFS and TFS. Further on, this is also in agreement with the report that CLL patients with a higher percentage of 13q deleted nuclei have a transcriptome pattern similar to del(11p) and del(17q) patients, abnormal miRNA expression, and dysregulated pathways implicated in apoptosis and proliferation (BCR and NF-kappaB signaling) [23]. Altogether, our data and previously published data support a chronology of molecular events involving first the del(13q) and afterwards, as the clone expends, important epigenetic modifications. However, one question remains to be answered which concerns the possible contribution of suppressor genes from the deleted 13q locus to the regulation of the methylation process. Moreover, in this locus, differentially methylated regions have been identified in CLL B cells compared to controls, and they are located within tumor suppressor genes including the deleted in lymphocytic leukemia (DLEU)1 and the DLEU2 variant ALT1 [20]. Further experiments are required to test whether these suppressor genes are in turn involved in the global DNA methylation process and whether DNA demethylation leads to an adverse clinical outcome per se or through subclonal driver mutations.

DNMT3A is thought to be critical in the control of DNA methylation in CLL, and such an assertion is supported by the observation that DNMT3A knockdown in rodent hematopoietic stem cells results in CLL development [24] and that DNMT3A is present in the top 1% of the under-expressed genes in human CLL B cells [25]. An interaction between the two CLL high-risk factors TCL1 and NOTCH1 with DNMT3A to inhibit its function is also reported in CLL B cells [26, 27]. Accordingly, we have investigated DNMTs in CLL B cells and observed that DNMT1 and DNMT3B are downregulated in the cytogenetic low- and intermediate/high-risk groups while DNMT3A reduction was restricted to CLLs carrying cytogenetic intermediate/high-risk factors. This result differs from previous publications, which have compared CLL B cells with peripheral blood normal B cells to show, in one case, normal DNMT1 expression, an upregulation of DNMT3A, and a reduction of DNMT3B in CLL B cells [28], and in the other case, researchers have reported a reduction in DNMT1 [10]. These discrepancies could be most likely due to the selection of the control cells as important variations in DNMTs are reported during B cell ontogeny and to the reference gene used for correction [29–31].

Despite its stability, 5-mCyt can be reversed firstly during DNA replication in a passive way or, secondly, by TET proteins which can mediate 5-mCyt oxidation to 5-hmCyt, and in a less efficient fashion to 5-formylcytosine (5-fCyt), and 5-carboxylcytosine (5-CaCyt). Next, transformation to an unmodified Cyt requires the coupled action of thymine DNA glycosylase (TDG) and the base excision repair (BER) system. TET1 and double TET1/TET2 knockdown promote B cell malignancies [32, 33]. First reported in myeloid lineage malignant diseases, 5-hmCyt level reduction was related to mutations in TETs or their regulators (e.g., IDH1/2, WT1) [34]. In CLL, DNA demethylation is observed, but mutations in TETs and IDHs are rare events. However, a reduction in TET1/3 and IDH2 was shown to be associated with a poor prognosis [12]. In this report, we could confirm that patients with low mRNA expression of TET1 and TET2 were predominantly classified into the cytogenetic intermediate/high-risk group, which is in agreement with our observation that both 5-mCyt and 5-hmCyt levels were reduced in the intermediate- and/or high-risk groups. Van Damme et al. observed a decreased expression of TET1 in CLL B cells but did not observe an association with DNA hydroxymethylation, and this may be explained, in part, by the small number of patients studied for hydroxymethylation levels [12]. In another study comparing CLL B cells to peripheral blood B cells, TET2 was reported to be overexpressed and such overexpression was not associated with other clinical and biological markers [35]. The list of the mechanisms controlling TET expression in leukemic cells is growing and involves microenvironmental factors [12], hypoxia, reactive oxygen species (ROS), and miRNAs such as miR-15a and miR-29 [36]. Additional experiments are required to better understand the mechanisms controlling active DNA demethylation in CLL and their consequences.

Relevant limitations of our study include the following: (i) global 5-mCyt and 5-hmCyt analysis is likely to reflect variations of repetitive elements (half of the genome) than gene regulatory sequence, which may render comparison with EWAS studies difficult; (ii) the absence of a well-established control B cell for CLL B cells, making difficult the comparison of the transcriptional analysis between normal and tumor cells; and (iii) a transcriptional analysis for DNMTs and TETs, which does not test enzyme activity.

In conclusion, our study established that loss in the 5-Cyt derivatives, 5-mCyt and, to a lesser extent, 5-hmCyt, during disease progression, better predict the CLL patients’ outcome when associated with the classical cytogenetic analysis. Future work is required to develop routine assays for monitoring DNA methylation during disease course and to test their utility as therapeutic response predictors.